Spaces:
Runtime error
Runtime error
File size: 10,717 Bytes
ee21b96 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 |
#!/usr/bin/env python3
"""Calculates the Frechet Inception Distance (FID) to evalulate GANs
The FID metric calculates the distance between two distributions of images.
Typically, we have summary statistics (mean & covariance matrix) of one
of these distributions, while the 2nd distribution is given by a GAN.
When run as a stand-alone program, it compares the distribution of
images that are stored as PNG/JPEG at a specified location with a
distribution given by summary statistics (in pickle format).
The FID is calculated by assuming that X_1 and X_2 are the activations of
the pool_3 layer of the inception net for generated samples and real world
samples respectivly.
See --help to see further details.
Code apapted from https://github.com/bioinf-jku/TTUR to use PyTorch instead
of Tensorflow
Copyright 2018 Institute of Bioinformatics, JKU Linz
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
"""
import os
from argparse import ArgumentParser, ArgumentDefaultsHelpFormatter
import numpy as np
import torch
import torch.utils.data
import torchvision.transforms as transforms
import tqdm
from PIL import Image
from scipy import linalg
from torch.nn.functional import adaptive_avg_pool2d
from torch.utils import data
from eval_utils.inceptionV3 import InceptionV3
class Dataset(data.Dataset):
'Characterizes a dataset for PyTorch'
def __init__(self, path, transform=None):
'Initialization'
self.file_names = self.get_filenames(path)
self.transform = transform
def __len__(self):
'Denotes the total number of samples'
return len(self.file_names)
def __getitem__(self, index):
'Generates one sample of data'
img = Image.open(self.file_names[index]).convert('RGB')
# Convert image and label to torch tensors
if self.transform is not None:
img = self.transform(img)
return img
def get_filenames(self, data_path):
images = []
for path, subdirs, files in os.walk(data_path):
for name in files:
if name.rfind('jpg') != -1 or name.rfind('png') != -1:
filename = os.path.join(path, name)
if os.path.isfile(filename):
images.append(filename)
return images
parser = ArgumentParser(formatter_class=ArgumentDefaultsHelpFormatter)
parser.add_argument('--batch-size', type=int, default=64,
help='Batch size to use')
parser.add_argument('--dims', type=int, default=2048,
choices=list(InceptionV3.BLOCK_INDEX_BY_DIM),
help=('Dimensionality of Inception features to use. '
'By default, uses pool3 features'))
parser.add_argument('-c', '--gpu', default='', type=str,
help='GPU to use (leave blank for CPU only)')
parser.add_argument('--path1', type=str, help='path to images')
parser.add_argument('--path2', type=str, help='path to images')
def get_activations(images, model, batch_size=64, dims=2048, cuda=False, verbose=True):
"""Calculates the activations of the pool_3 layer for all images.
Params:
-- images : Numpy array of dimension (n_images, 3, hi, wi). The values
must lie between 0 and 1.
-- model : Instance of inception model
-- batch_size : the images numpy array is split into batches with
batch size batch_size. A reasonable batch size depends
on the hardware.
-- dims : Dimensionality of features returned by Inception
-- cuda : If set to True, use GPU
-- verbose : If set to True and parameter out_step is given, the number
of calculated batches is reported.
Returns:
-- A numpy array of dimension (num images, dims) that contains the
activations of the given tensor when feeding inception with the
query tensor.
"""
model.eval()
# d0 = images.shape[0]
d0 = images.__len__() * batch_size
if batch_size > d0:
print(('Warning: batch size is bigger than the data size. '
'Setting batch size to data size'))
batch_size = d0
n_batches = d0 // batch_size
n_used_imgs = n_batches * batch_size
pred_arr = np.empty((n_used_imgs, dims))
# for i in range(n_batches):
for i, batch in tqdm.tqdm(enumerate(images)):
# batch = batch[0]
# if verbose:
# print('\rPropagating batch %d/%d' % (i + 1, n_batches), end='', flush=True)
# import ipdb
# ipdb.set_trace()
start = i * batch_size
end = start + batch_size
# batch = torch.from_numpy(images[start:end]).type(torch.FloatTensor)
# batch = Variable(batch, volatile=True)
if cuda:
batch = batch.cuda()
pred = model(batch)[0]
# If model output is not scalar, apply global spatial average pooling.
# This happens if you choose a dimensionality not equal 2048.
if pred.shape[2] != 1 or pred.shape[3] != 1:
pred = adaptive_avg_pool2d(pred, output_size=(1, 1))
pred_arr[start:end] = pred.cpu().data.numpy().reshape(batch_size, -1)
if verbose:
print(' done')
return pred_arr
def calculate_frechet_distance(mu1, sigma1, mu2, sigma2, eps=1e-6):
"""Numpy implementation of the Frechet Distance.
The Frechet distance between two multivariate Gaussians X_1 ~ N(mu_1, C_1)
and X_2 ~ N(mu_2, C_2) is
d^2 = ||mu_1 - mu_2||^2 + Tr(C_1 + C_2 - 2*sqrt(C_1*C_2)).
Stable version by Dougal J. Sutherland.
Params:
-- mu1 : Numpy array containing the activations of a layer of the
inception net (like returned by the function 'get_predictions')
for generated samples.
-- mu2 : The sample mean over activations, precalculated on an
representive data set.
-- sigma1: The covariance matrix over activations for generated samples.
-- sigma2: The covariance matrix over activations, precalculated on an
representive data set.
Returns:
-- : The Frechet Distance.
"""
mu1 = np.atleast_1d(mu1)
mu2 = np.atleast_1d(mu2)
sigma1 = np.atleast_2d(sigma1)
sigma2 = np.atleast_2d(sigma2)
assert mu1.shape == mu2.shape, \
'Training and test mean vectors have different lengths'
assert sigma1.shape == sigma2.shape, \
'Training and test covariances have different dimensions'
diff = mu1 - mu2
# Product might be almost singular
covmean, _ = linalg.sqrtm(sigma1.dot(sigma2), disp=False)
if not np.isfinite(covmean).all():
msg = ('fid calculation produces singular product; '
'adding %s to diagonal of cov estimates') % eps
print(msg)
offset = np.eye(sigma1.shape[0]) * eps
covmean = linalg.sqrtm((sigma1 + offset).dot(sigma2 + offset))
# Numerical error might give slight imaginary component
if np.iscomplexobj(covmean):
if not np.allclose(np.diagonal(covmean).imag, 0, atol=1e-3):
m = np.max(np.abs(covmean.imag))
raise ValueError('Imaginary component {}'.format(m))
covmean = covmean.real
tr_covmean = np.trace(covmean)
return (diff.dot(diff) + np.trace(sigma1) +
np.trace(sigma2) - 2 * tr_covmean)
def calculate_activation_statistics(images, model, batch_size=64,
dims=2048, cuda=False, verbose=True):
"""Calculation of the statistics used by the FID.
Params:
-- images : Numpy array of dimension (n_images, 3, hi, wi). The values
must lie between 0 and 1.
-- model : Instance of inception model
-- batch_size : The images numpy array is split into batches with
batch size batch_size. A reasonable batch size
depends on the hardware.
-- dims : Dimensionality of features returned by Inception
-- cuda : If set to True, use GPU
-- verbose : If set to True and parameter out_step is given, the
number of calculated batches is reported.
Returns:
-- mu : The mean over samples of the activations of the pool_3 layer of
the inception model.
-- sigma : The covariance matrix of the activations of the pool_3 layer of
the inception model.
"""
act = get_activations(images, model, batch_size, dims, cuda, verbose)
mu = np.mean(act, axis=0)
sigma = np.cov(act, rowvar=False)
return mu, sigma
def _compute_statistics_of_path(path, model, batch_size, dims, cuda):
if path.endswith('.npz'):
f = np.load(path)
m, s = f['mu'][:], f['sigma'][:]
f.close()
else:
dataset = Dataset(path, transforms.Compose([
transforms.Resize((299, 299)),
transforms.ToTensor(),
]))
print(dataset.__len__())
if dataset.__len__() < batch_size:
batch_size = 1
dataloader = torch.utils.data.DataLoader(dataset=dataset, batch_size=batch_size, shuffle=False, drop_last=True,
num_workers=0)
m, s = calculate_activation_statistics(dataloader, model, batch_size, dims, cuda)
return m, s
def calculate_fid_given_paths(paths, batch_size, cuda, dims):
"""Calculates the FID of two paths"""
for p in paths:
if not os.path.exists(p):
raise RuntimeError('Invalid path: %s' % p)
block_idx = InceptionV3.BLOCK_INDEX_BY_DIM[dims]
model = InceptionV3([block_idx])
if cuda:
model.cuda()
m1, s1 = _compute_statistics_of_path(paths[0], model, batch_size, dims, cuda)
m2, s2 = _compute_statistics_of_path(paths[1], model, batch_size, dims, cuda)
fid_value = calculate_frechet_distance(m1, s1, m2, s2)
return fid_value
if __name__ == '__main__':
args = parser.parse_args()
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
paths = ["", ""]
paths[0] = args.path1
paths[1] = args.path2
print(paths)
fid_value = calculate_fid_given_paths(paths, args.batch_size, args.gpu, args.dims)
print('FID: ', fid_value)
|